This lesson will familiarizing the student with cloud and deployment terminology along with demonstrating how deployment fits within the machine learning workflow.
Building a Model using SageMaker
Learn how to use Amazon's SageMaker service to predict Boston housing prices using SageMaker's built-in XGBoost algorithm.
Deploying and Using a Model
In this lesson students will learn how to deploy a model using SageMaker and how to make use of their deployed model with a simple web application.
Hyperparameter Tuning
In this lesson students will see how to use SageMaker's automatic hyperparameter tuning tools on the Boston housing prices model from lesson 2 and with a sentiment analysis model.
Updating a Model
In this lesson students will learn how to update their model to account for changes in the underlying data used to train their model.
Train and deploy unsupervised models(PCA and k-means clustering) to group US counties by similarities and differences. Visualize the trained model attributes and interpret the results.
Payment Fraud Detection
Train a linear model to do credit card fraud detection. Improve the model by accounting for class imbalance in the training data and tuning for a specific performance metric.
Interview Segment: SageMaker as a Tool & the Future of ML
If you're interested in how SageMaker has developed to serve businesses and learners, take a look at this short interview segment with Dan Mbanga.
Deploying Custom Models
Design and train a custom PyTorch classifier by writing a training script. This is an especially useful skill for tasks that cannot be easily solved by built-in algorithms.
Time-Series Forecasting
Learn how to format time series data into context(input) and prediction(output) data, and train the built-in algorithm, DeepAR; this uses an RNN to find recurring patterns in time series data.
Convolutional Neural Networks allow for spatial pattern recognition. Alexis and Cezanne go over how they help us dramatically improve performance in image classification.
GPU Workspaces Demo
See a demonstration of GPU workspaces in the Udacity classroom.
Cloud Computing
Take advantage of Amazon's GPUs to train your neural network faster. In this lesson, you'll setup an instance on AWS and train a neural network on a GPU.
Transfer Learning
Learn how to apply a pre-trained network to a new problem with transfer learning.
Weight Initialization
In this lesson, you'll learn how to find good initial weights for a neural network. Having good initial weights can place the neural network closer to the optimal solution.
Autoencoders
Autoencoders are neural networks used for data compression, image denoising, and dimension reduction. Here, you'll build autoencoders using Pytorch.